The star formation history of mass-selected galaxies in the COSMOS field

We explore the evolution of the specific star formation rate (SSFR) for 3.6um-selected galaxies of different M_* in the COSMOS field. The average SFR for sub-sets of these galaxies is estimated with stacked 1.4GHz radio continuum emission. We separately consider the total sample and a subset of galaxies (SF) that shows evidence for substantive recent star formation in the rest-frame optical SED. At 0.2<z<3 both populations show a strong and M_*-independent decrease in their SSFR towards z=0.2, best described by a power- law (1+z)^n, where n~4.3 for all galaxies and n~3.5 for SF sources. The decrease appears to have started at z>2, at least above 4x10^10M_Sun where our conclusions are most robust. We find a tight correlation with power-law dependence, SSFR (M_*)^beta, between SSFR and M_* at all z. It tends to flatten below ~10^10M_Sun if quiescent galaxies are included; if they are excluded a shallow index beta_SFG -0.4 fits the correlation. On average, higher M_* objects always have lower SSFRs, also among SF galaxies. At z>1.5 there is tentative evidence for an upper SSFR-limit that an average galaxy cannot exceed. It is suggested by a flattening of the SSFR-M_* relation (also for SF sources), but affects massive (>10^10M_Sun) galaxies only at the highest z. Below z=1.5 there thus is no direct evidence that galaxies of higher M_* experience a more rapid waning of their SSFR than lower M_* SF systems. In this sense, the data rule out any strong 'downsizing'. We combine our results with recent measurements of the galaxy (stellar) mass function in order to determine the characteristic mass of a SF galaxy (M_*=10^(10.6\pm0.4)M_Sun). In this sense, too, there is no 'downsizing'. Our analysis constitutes the most extensive SFR density determination with a single technique to z=3. Recent Herschel results are consistent with our results, but rely on far smaller samples.